Putting algorithmic bias on top of the agenda in the discussions on autonomous weapons systems

<jats:title>Abstract</jats:title><jats:p>Biases in artificial intelligence have been flagged in academic and policy literature for years. Autonomous weapons systems—defined as weapons that use sensors and algorithms to select, track, target, and engage targets without human intervention—have the potential to mirror systems of societal inequality which reproduce algorithmic bias. This article argues that the problem of engrained algorithmic bias poses a greater challenge to autonomous weapons systems developers than most other risks discussed in the Group of Governmental Experts on Lethal Autonomous Weapons Systems (GGE on LAWS), which should be reflected in the outcome documents of these discussions. This is mainly because it takes longer to rectify a discriminatory algorithm than it does to issue an apology for a mistake that occurs occasionally. Highly militarised states have controlled both the discussions and their outcomes, which have focused on issues that are pertinent to them while ignoring what is existential for the rest of the world. Various calls from civil society, researchers, and smaller states for a legally binding instrument to regulate the development and use of autonomous weapons systems have always included the call for recognising algorithmic bias in autonomous weapons, which has not been reflected in discussion outcomes. This paper argues that any ethical framework developed for the regulation of autonomous weapons systems should, in detail, ensure that the development and use of autonomous weapons systems do not prejudice against vulnerable sections of (global) society.</jats:p>